基于差分理论的短期负荷预测神经网络模型

Short Term Load Forecasting Using Neural Networks Based On Difference Theory

电力负荷是受周期性变化以及天气等因素影响的高度非线性系统,而神经网络仅仅对已学习过的模式具有较好的范化能力。为提高神经网络的负荷预测精度,提出先对原始负荷序列进行差分运算以除去其周期性影响,然后依据相似性原理建立RBF神经网络预测模型,仿真实验表明采用该方法短期负荷预测精度有所改善。

Electric load is a highly nonlinear system, which is affected by the cyclical changes and other factors such as weather changes. The neural networks have better generalization only to the studied models. In order to improve the forecasting performance,a new method based on neural networks is proposed in this paper. First the difference operation is applied to the original load time se-ries in order to diminish the cyclical influence, then a RBF neural network forecasting model is established based on the principle of similarities. Numerical experiments demonstrate that the precision is improved with the model proposed in this paper.